Source: The Armchair Economist, Economics and Everyday Life, Steven E. Landsburg, The Free Press, New York, 1995, p. 211-219






How Economists Go Wrong


Once there was an economist who wanted to understand football. He knew the rules but had no feeling for the game. So he decided to observe the great coaches and to learn from them.

Each time he watched a game, the economist painstakingly recorded all of the plays that were called and all of the surrounding circumstances that might have been relevant. Each night he sophisticated statistical tests to reveal hidden patterns in the data. Eventually his research began to pay off. He discovered that quarterbacks often throw the ball in the direction of a receiver, that the ball carrier usually runs in the direction of the opposing team's goalpost, and that field goals the final minute are most often attempted by teams that are or two points behind.

One day the commissioner of the National Football League became concerned about punting. He had come to believe that teams punt far too often, and that their behaviour is detrimental to the game. (Exactly why he thought this has never been determined, but he was quite sure of himself.) The commissioner became obsessed with the need to discourage punting called in his assistants for advice on how to cope with the problem.

One of those assistants, a fresh M.B.A., breathlessly announced that he had taken courses from an economist who was a great expert on all aspects of the game and who had developed detailed statistical models to predict how teams behave. He proposed retaining the economist to study what makes teams punt.

The commissioner summoned the economist, who went home with a large retainer check and a mandate to discover the causes of punting. Many hours later (he billed by the hour) the answer was at hand. Volumes of computer printouts left no doubt: Punting nearly always takes place on the fourth down.

But the economist was trained in the scientific method and knew that describing the past is less impressive than predicting the future. So before contacting the commissioner, he put his model to the acid test. He attended several football games and predicted in advance that all punting would take place on fourth down. When his predictions proved accurate, he knew he had made a genuine scientific discovery.

The commissioner, however, was not paying for pure science. Knowledge for its own sake might satisfy a philosopher, but the commissioner had a practical problem to solve. His goal was not to understand punting but to eradicate it.

So the commissioner sent the economist back to his computers to formulate a concrete policy proposal. After a few false starts, the economist had a brainstorm. What if teams were allowed only three downs?

To test his idea, the economist wrote a computer program to simulate the behaviour of teams in a game with three downs. The program was written to fully incorporate everything the economist knew about when teams punt. Simulation after simulation confirmed his expectation: Because punting takes place on fourth down only, nobody punts in a game without fourth downs.

The commissioner was impressed by the weight of the evidence and held a press conference to announce a change in the rules of football. From now on, only three downs would be permitted. The commissioner announced his confidence that the days of excessive punting were behind us. But the reality was otherwise. Teams began punting on third down, and the commissioner stopped listening to economists.

Our hero was well within the mainstream of twentieth-century policy analysis. In the years following World War II, economists learned statistics. The new subject of econometrics made it possible to detect deep patterns in economic data and to test whether those patterns were likely to be repeated. Economists scrutinised consumption behaviour, investment decisions, farm output, labour supply, sales of financial assets, and everything else they could think of. And the enterprise succeeded beyond their dreams. The data revealed striking consistencies that were used to predict the future with remarkable accuracy.

A contemporary American might find it difficult to imagine a time when macroeconomic predictions were frequently correct. But that brief golden era did exist. The natural question is, What went wrong?

What went wrong appears to be that governments started taking economists seriously, and that this development undermined everything. Let us follow the trail of one particular economist, formerly a consultant to the National Football League, and now employed by the U.S. government to help formulate economic policy.

The goal was to stimulate agricultural production. Our hero was assigned to analyse the cereal market and design a policy that would put more corn flakes on the average American breakfast table.

The first task was to determine the facts about corn flake consumption. After many months of poring over data, the economist found the statistical regularity he was looking for. The average family buys two boxes of corn flakes every month. This behaviour is remarkably consistent. For example, small changes in after-tax income have almost no effect on corn flake sales.

Ever the skeptical scientist, the economist was unwilling to rely exclusively on historical data. Instead, he put his theory to the acid test of prediction. He forecast that over the next several months, families would continue buying about two boxes of corn flakes every month regardless of small fluctuations in income. His forecasts were repeatedly confirmed. His sense of triumph recalled that glorious day in his youth when he had first detected the fourth down/punting connection.

The economist's superiors were pleased with his finding, and even more pleased when he made it the basis of a policy proposal: Let the government provide each American family with two boxes of corn flakes every month. Financing the program will require a small tax increase, but we know that small tax increases don't affect corn flake sales. Therefore families will continue buying two boxes a month at the grocery store. Together with the two boxes that the government gives them, they will consume a total of four boxes, or twice as much as they used to.

But a strange thing happened. When the government started giving away corn flakes, shoppers reacted like football players given only three downs to gain ten yards: They changed their strategies. As soon as people realised that the government was delivering corn flakes to their doorsteps, they stopped buying corn flakes at the grocery stores.

Our economist-hero is no exaggerated fiction but a true representative of his generation. In the 1950s and 60s, his path was the path to fame and glory. Only 20 years ago, Robert E. Lucas, Jr. (now of the University of Chicago), issued the first widely recognised warning that human beings respond to policy changes, and that this simple observation renders traditional policy analysis completely invalid. Even today, college students taking their first economics course are taught to assume that when the government provides corn flakes, people go on buying corn flakes just as before. (Of course, the textbooks express this assumption in terms of algebra rather than corn flakes, to insure that students will not understand what it means.)

Unfortunately for policy analysts, people are not simple automatons. They are strategic players in a complicated game where government policies set some of the rules. The behaviours that economists can observe - the decision to buy a car or a house, to quit one job or to take a new one, to hire additional workers or to build a new factory - are bits of strategy. As long as the rules stay fixed, we can reasonably expect the strategies not to change very much, and we can accurately extrapolate from past observations. When the rules change, all bets are off. Our economist-hero would have been well advised to devote less effort to his statistics and more to pure theory. Guided by the right theory of football - which is that each team attempts to score more points than the other - he could have accurately predicted how players would respond to a new set of rules. Guided by the right theory of corn flakes - which is that families decide how much to eat on the basis of taste, convenience, price, and available alternatives - he could have accurately predicted that letting the government do people's shopping would not make them any hungrier.

Of course, some theories are wrong, and economists who subscribe to those theories do not predict accurately. But an economist with a theory has at least a chance that his theory is the right one. An economist who relies on nothing but statistical extrapolations has no chance at all.

The area where macroeconomists have failed most spectacularly is in the relationship between employment and inflation. For many years, good evidence indicated a powerful correlation: Times of high inflation are times of high employment and vice versa. By the late 1960s, this observation had survived rigorous statistical testing and was generally accepted as a scientific truth. Accepting that truth as a basis for policy, politicians attempted to manipulate the inflation rate as a means of controlling unemployment. The result was contrary to all expectations: a decade of stagflation - high inflation and low employment combined. Then in the 1980's, inflation fell dramatically, and, after an initial severe recession, employment opportunities expanded at unprecedented rates. The old statistical regularities seemed to have been turned on their heads.

What had changed? It is impossible to answer that question without a theory of how the inflation rate affects individual employment decisions. In 1971, Robert Lucas offered the first I example of such a theory.

Imagine Willie Worker, currently unemployed not because he has literally no job opportunities, but because his opportunities are so unattractive that he prefers unemployment. Willie's best wage offer is $10,000 a year, which would barely cover the cost of getting to work. If the wage were $15,000, Willie would take the job.

One night, while Willie sleeps, there is a massive inflation, causing all prices and all wages to double. The employer who offered $10,000 yesterday offers $20,000 today. That's still not good enough, though. In a world of doubled prices, Willie doesn't want to work for less than $30,000. He remains unemployed.

Now let me change the story only slightly. The morning after the night of the great inflation, Willie is awakened by a phone call from an employer offering $20,000. At this point, Willie has not yet read the morning papers and is unaware that prices have changed. He happily reports for work. Only on his way home, stopping at the supermarket to spend his first paycheck, does Willie discover the cruel truth and begin composing a letter of resignation.

This highly stylised fable captures a potentially important aspect of reality. One way that inflation can increase employment is by fooling people. It makes job opportunities look more attractive than they usually are and entices workers to accept jobs they would reject if they knew more about the economic environment.

We can tell pretty much the same story from the employer's viewpoint. Suppose you own an ice cream parlor, selling ice cream cones at one dollar apiece. If you could sell them for two dollars apiece, you would expand your operation, but you've learned by experimenting that two dollars is more than the traffic will bear.

If all prices and wages - including all of your costs - were to double, then you'd be able to sell cones for two dollars, but that two dollars would be worth no more than one dollar was worth yesterday. You would continue as before.

But suppose that prices and wages double without your noticing. You notice only that your customers suddenly seem willing to pay more for their ice cream cones. (Probably you first discover this when traffic picks up, because your one-dollar cones have begun to seem like quite a bargain to customers whose wages have doubled.) You expand your operation and hire a lot of new workers. Even after you discover your mistake, part of the expansion is irrevocable: The new freezers are in place, the new parking spaces are under construction, and you might want to keep at least some of those new employees.

The Lucas story implies not that inflation puts people to work but that unanticipated inflation puts people to work. In his story, fully anticipated inflations do not affect anyone's behaviour. A (highly stylised) history of modem macroeconomics would go something like this: Inflations fool workers into accepting more jobs and employers into hiring more workers. Governments notice that inflation is consistently accompanied by high employment and decide to take advantage of this relationship by systematically manipulating the inflation rate. Workers and employers quickly notice what the government is up to and cease to be fooled. The correlation between inflation and unemployment breaks down precisely because the government attempts to exploit it.

Let me be entirely explicit about the analogy. Throughout the history of football, there has been no distinction between fourth downs and last downs. If economist A asserts that "teams punt only on fourth down" and economist B asserts that "teams punt only on last down," then nothing in the historical data can distinguish between their hypotheses. Anything that goes to confirm economist A's theory will go to confirm economist B's theory, and vice versa. Both theories will predict equally accurately until the rules change. But after the rules change, when the last down becomes the third down instead of the fourth, one theory will continue to be correct while the other goes drastically wrong.

Throughout the history of corn flakes, there has been no distinction between corn flakes purchased and corn flakes eaten. If economist A asserts that "families purchase two boxes of corn flakes per month" and economist B asserts that "families eat two boxes of corn flakes per month, then nothing in the historical data can distinguish between their hypotheses. Anything that goes to confirm economist A's theory will go to confirm economist B's theory, and vice versa. Both theories will predict equally accurately until the rules change. But after the rules change, when the government provides each family with two boxes of corn flakes over and above what they purchase for themselves, one theory will continue to be correct while the other goes drastically wrong.

Throughout the two decades following World War II, fluctuations in the inflation rate were largely unanticipated. There was no distinction between inflation and unanticipated inflation. If economist A asserts that inflation puts people to work and economist B asserts that unanticipated inflation puts people to work, then nothing in the historical data can distinguish between their hypotheses. Anything that goes to confirm economist A's theory will go to confirm economist B's theory, and vice versa. Both theories will predict equally accurately until the rules change. But after the rules change, when the government starts systematically manipulating the inflation rate in foreseeable ways, one theory will continue to be correct while the other goes drastically wrong.

With nothing but history as a guide, predicting human behaviour in a fixed environment is easy; predicting human behaviour in a changing environment is impossible. In New York in the summertime, I carry an umbrella to work if the morning sky is mostly gray. If you watched me for a while, you would probably notice this pattern and get good at predicting when I was going to carry an umbrella. But in Colorado in the summertime, I never carry an umbrella to work, because it is virtually certain that the regular afternoon thunderstorm will be over before I leave the office at 5:00. Move me to Colorado and your predictions will go completely haywire.

An economist who understands why teams punt knows what will happen if you change the rules; an economist who understands why people buy cereal knows what will happen if you give out free corn flakes; an economist who understands why people accept certain job offers knows what will happen if you manipulate the inflation rate; and an economist who understands why I carry an umbrella knows what will happen if I relocate to a desert. To understand behaviour, economists must tell stories - stories like the tale of the unemployed worker or the saga of the ice cream parlor - and spend a lot of time worrying about whether their stories are plausible, and how they can tell better ones.

Many economists are deeply unsatisfied with the Lucas stories and ask embarrassing questions like "Why can't the ice cream store owner learn the inflation rate from the Wall Street Journal before he embarks on a massive expansion program?" In response, Lucas and others have constructed increasingly elaborate versions of the original story, and also a host of competing stories.

But whatever the fate of any particular story, Lucas permanently changed macroeconomics by his insistence that a macroeconomist must have some story to tell and must tell it in sufficient detail so that its flaws are readily apparent. In 1971, Lucas began his paper on "Expectations and the Neutrality of Money" by describing all of the minutiae of an artificial society, including the life spans of its citizens, the age at which they retire, and exactly how much they can observe of each others' private affairs. Given those precise specifications, he was able to trace every consequence of an increase in the money supply. In the Lucas world, a random fluctuation in the money supply increases both inflation and employment. That same fluctuation, when it occurs not randomly but as a component of government policy, increases inflation but leaves employment unchanged. According to legend, when Lucas submitted his paper to one of the premier economics journals, the rejection letter suggested that the paper was interesting but bore no relation to macroeconomics. Today, that paper is the archetype that defines what macroeconomics is about. Some economists like the story and others hate it, but there is a widespread consensus that our best hope is tell and dissect explicit stories about worlds that are simple enough to understand, but complicated enough to bear some relation to the world we inhabit. That is a radical departure from the old macroeconomics, and a necessary one. As a predictive science, modern macroeconomics has yet to succeed. But modern macroeconomics is only 20 years old, determined not to repeat the mistakes of its elders, and eyeing the future with the impatient confidence of youth.